There have been several papers recently addressing questions of niche overlap, niche shifts, etc. In all cases these authors have used models (mostly correlative species distribution models) built using distribution data in the native or invaded ranges. My question deals with the issue of whether the model is the niche, or whether the niche is really something resulting from the interaction between the taxon's genome and its environment, which is apparent only through observation of the effects of its interaction with the environment. In the case of species distribution records, the observations are captured haphazardly typically in a biased, inaccurate and error-prone manner. When challenged to think about the question, I think most scientists would agree that a species niche is really the emergent ecological pattern, and the model is an abstract approximation (oftentimes reflecting the errors biases etc.). So what?! The importance here is that the formulation of the SDM's involves selection of a set of covariates (usually bioclim variables) and relating them somehow to the distribution data. The selection of the covariates and the model structure have demonstrable impacts on the resulting models, and sometimes result in the production and publication of absurd models supported by inappropriate measures of the goodness of fit. So, my question is really about whether these niche comparison papers are over-stating the implied precision and confidence levels in their analyses, and perhaps drawing inappropriate conclusions. If we accept that the model is not the niche, but only an approximation of it, then if we compare two approximations (both involving extrapolation of correlative regression models) we should probably exhibit more circumspection in the text describing the results and the consequent discussion than we have seen to date.
Dear Darren, your concerns are mine as well; both on the conceptual and empirical level. Therefore, I stay in my articles close to the empirical and avoid "habitat" and "niche"; instead, I use occupied range, biophysical range (fundamental niche?), summer range (summer habitat?) and the likes; further, I try to model plants, animals and ecosystems with a fundamental set of environmental co-variates/variables representing major ecosystem components (bio-climate, substrate, vegetation, land management). And yes, using bioclim only and uncritically is hardly scientific; I am always surprised to see how critical authors can be about species records and uncritical in the same paper about the biased meteo-station locations on which bioclim is based (see attached for an example). Not only are meteo-station locations always biased, the selection out of the stations is rarely transparent; further, a closer look at meteorological data series often shows data gaps, change of instrumentation and calculation (e.g. from snow to precipitation) over the decades. Similar considerations apply to other environmental variables. In other words, we need to be as critical and transparent about the independent variables as we are about the independent variable in SDM.
Article Where the bears roam in Majella National Park, Italy
Article Fine resolution distribution modelling of endemics in Majell...
I do not think any model could represent the true "niche" of a species, since models invariably have error associated with them. I'm not as versed in the models about which you write, but biologically there is a further layer of complexity in whether you mean the fundamental niche vs realized niche. The fundamental niche is the niche that theoretically could be filled by the organism, while the realized niche is what the organism occupies given interspecific interactions (e.g. maybe a species that tolerates fresh- and saltwater theoretically could be found in a freshwater environment, but if the freshwater environment has superior competitors, it may only in nature be able to occupy saltwater). Perhaps the models are better at assessing a fundamental niche than observations in nature would be. I think scientists sometimes use terminology a bit loosely though, and sometimes the best approach is to just explicitly define controversial terms in the article.
Some notes:
1) "Species niche" is a mistake, a false concept. Really, a species is a subjective concept, a thing about which a taxonomist means that it is a species. Really, the life is a complex of many populations, not of species. There are exist many species consisting of ecologically different populations, and these differences often are based on different gene pools.
2) It is impossible, to describe all interspecific and even all interpopulational interactions using Hutchinsonian tolerance niche models. At least three causes exist. 1) Typical multidimensional niche models don't take into account the feedbacks between a population and it's environment; 2) if we try to view interpopulational ("interspecific") interactions inside the bounds of a community or a homogeneous ecosystem, we forget that most populations are represented in more than one community. Some cases are obvious: that are e.g. migratory birds and other "heterocoene" forms. Some other examples are not so well-known although also are typical the "heterotopic" forms which are moving between different communities (e.g. sea birds). I think, the mismatch of populations boundaries and communities boundaries is a general life cover feature. 3) Any typical population is ecologically heterogeneous, it consists of several ecologically different functional parts (econes). Therefore also the niche cannot be a simple hypervolume in the Hutchinsonian hyperspace, is should be a very complex structure
Hi Darren. I am overly practical so I won't probably give you answers you need for modelling. And I probably have not grasped what you are after in an answer. But I suspect that niche modelling in largely okay but gets scratchy as one enters the edge of the ecological zone. For Qfly, we know they prefer certain sites for survival (summer and winter) and different ones for mating. The model may well predict a favourable climate model but is compromised by microclimate effects which I suspect a model can't do. Additionally modelling is fine using a great collection of weather stations but if there are no susceptible hosts, the modelling will create a false potential for survival so there needs to be some ground trurthing for hosts. That is very costly compared to computer modelling. I happen to be reading your 2010 cucmis paper last night - we are doing a review of cucumis to assist with market access. It can all get very cofusing. Anyway that is my two-bobs worth. All the best to you and Tania.
Niche models are only as good as the data. Common practice, as you said, is to examine physioclimate variables like rainfall, elevation, temperature. However, species also have physiographic/structural requirements that, to the best of my knowledge, are ... overlooked because these data take a lot of effort to obtain. Should desire to make use of these niche models to predict species response to climate change, invasiveness, then the state of the art is wanting and I'm left wondering why to bother.
There are many definition of a species niche. Species niche may mean the role and position of a species within the ecosystem in terms of how to meet needs for food, and shelter
It is better to talk about habitat suitability and not about niche, since distribution is not only restricted by bioclimatic conditions. Many times mutualistic or antagonic relationships are stronger explanations, as usually depending on the species we are studying.
Reading the previous responses helps me see how much more complicated this problem is than even the exceedingly practical aspects/questions I have noticed re my activism in advocating assisted migration for even common forest trees in this rapidly changing climate. Fundamentally, for slow maturing, large-seeded trees, we cannot just look at where the natural distribution actually was recorded at the baseline date of 1971 and then attempt to find among the 23 possible habitat/climate factors the ones that, when inputted into the model would fairly well lay out the 1971 range. My concern is that even in 1971 many of our common forest trees had not kept pace with climate change, and thus the northernmost locations of their mapped ranges really didn't indicate the ecological/physiological limits for them. Why? For trees in the mountain west, once a population starts heading uphill rather than north across the flats to to the next range, it cannot reverse track and undertake that journey as climate continues to warm. And for trees (e.g. oaks and beeches) in the east, once the Passenger Pigeon went extinct, it was game over for quick, quirky, long-distance dispersal. And yet I find those maps (e.g., the ones produced by USDA/Forest Service for 76 species out west and an equivalent multi-agency website for 134 trees species in eastern USA -- Climate Change Tree Atlas -- absolutely vital for giving me any sort of baseline at all. We need to begin with something, after all. Both websites and the papers underlying them always acknowledge that the modeled future habitat shifts in accordance with various accepted climate models show future range maps that have absolutely nothing to say about whether the trees can actually get to those places by the decade used in the projection. Indeed, reading the papers, I see the researchers sometimes use words such as "dire" to caution about how difficult it will be for those tree species to actually occupy that range on their own accord. Hence top forest researchers in USA and Canada are very comfortable with the idea of assisted migration. To learn more, check out the annotated list of papers I show in the forestry section of my assisted migration scholarly papers webpage: http://www.torreyaguardians.org/assisted-migration.html#forestry
And then if anyone has a response/advice, I'd be grateful to hear. What you all are trying to figure out has a hugely important use for those of us actively engaged in forestry conservation in this time of rapid climate change. (Note, I am the founder of the activist group, Torreya Guardians. Look us up online.
If you have good data, proceed with niche modeling. interpreting natural distribution is depend on how the data incorporated in your model.
- Badges
- Science topic
- Similar topics
- Biological Science
- Evolutionary Biology